Escolar Documentos
Profissional Documentos
Cultura Documentos
Nonlinear Opt
5. Nonlinear Optics
5.4.6
5.4.7
5.4.8
5.4.9
Nonlinear Polarization
and Nonlinear Susceptibilities ...............
5.2
5.7
5.3
5
5
8
9
10
10
11
13
5.1
13
14
16
17
22
22
22
24
25
27
29
29
31
33
37
38
40
43
45
46
46
Part A 5
Part A
49
53
53
54
55
57
60
Part A 5
63
63
64
66
69
71
71
71
74
78
80
1963 provided 20%30% efficiency of frequency conversion [5.5, 6]. The early phases of the development
and the basic principles of nonlinear optics have been
reviewed in the most illuminating way in the classical books by Bloembergen [5.7] and Akhmanov and
Khokhlov [5.8], published in the mid 1960s.
Over the following four decades, the field of nonlinear optics has witnessed an enormous growth, leading
to the observation of new physical phenomena and giving rise to novel concepts and applications. A systematic
introduction into these effects along with a comprehensive overview of nonlinear-optical concepts and devices
can be found in excellent textbooks by Shen [5.9],
Boyd [5.1], Butcher and Cotter [5.10], Reintjes [5.11]
and others. One of the most recent up-to-date reviews of
the field of nonlinear optics with an in-depth discussion
of the fundamental physics underlying nonlinear-optical
interactions was provided by Flytzanis [5.12]. This
chapter provides a brief introduction into the main
nonlinear-optical phenomena and discusses some of the
most significant recent advances in nonlinear optics, as
well as novel applications of nonlinear-optical processes
and devices.
Nonlinear Optics
E=
(5.1)
(5.2)
(5.3)
t
J( ) d ,
(5.5)
(5.6)
(P Q) + c ( M) ,
t
where P and Q are the electric dipole and electric
quadrupole polarizations, respectively. In the electric
dipole approximation, we keep only the first term on
the right-hand side of (5.6). In view of (5.5), this gives
the following relation between the D, E, and P vectors:
J=
(5.7)
(5.8)
and
(5.10)
PL = PL () exp ikr t + c.c. ,
(5.11)
(5.12)
(1) () =
(5.13)
(5.14)
(3)
where
and
are the second- and third-order
nonlinear susceptibilities.
Representing the electric field in the form of a sum
of plane monochromatic waves,
Ei (i ) exp(iki r i t) + c.c. ,
(5.15)
E=
i
(5.16)
Part A 5.1
D = E + 4 P.
P = PL + PNL ,
(5.4)
Part A
where
(5.17)
P (2) () = (2) (; i , j ) : E(i )E( j ) ,
.
P (3) () = (3) (; i , j , k )..E(i )E( j )E(k ) ,
1, k 1
(2)
(5.18)
(2) (; i , j ) = (2) ( = i + j )
=
(2) (t1 , t2 ) exp[i(i t1 + j t2 )] dt1 dt2 (5.19)
is the second-order nonlinear-optical susceptibility and
(3) (; i , j , k ) = (3) ( = i + j + k )
=
(3) (t1 , t2 , t3 )
exp[i(i t1 + j t2 + k t3 )] dt1 dt2 dt3
3, k 3
2, k 2
Fig. 5.1 Sum-frequency generation 1 + 2 = 3 in a medium with a quadratic nonlinearity. The case of 1 = 2
corresponds to second-harmonic generation
(2)
(5.20)
Part A 5.2
(5.22)
(5.21)
(5.23)
where k and A(z, t) are the wave vector and the envelope
of the electric field, k p and PNL (z, t) are the wave vector
and the envelope of the polarization wave.
If the envelope A(z, t) is a slowly varying function over the wavelength, | 2 A/z 2 | |kA/z|, and
2 PNL /t 2 2 PNL , (5.21) is reduced to [5.9]
A 1 A 2i2
+
=
PNL exp (ikz) ,
z u t
kc2
(5.24)
Nonlinear Optics
where
1 =
221 (2)
(; 2, )
k1 c2
2 =
421 (2)
k2 c2 SHG
(5.27)
and
Lc2 = 2Lc1
0.4
0.3
(5.28)
0.2
Lc1
0.1
0.0
6
L /Lc
length L of the nonlinear medium normalized to the coherence length L c for two values of L c : (dashed line) L c1 and
(solid line) L c2 = 2L c1
Part A 5.3
1 A1
A1
+
= i1 A1 A2 exp (ikz) ,
z
u 1 t
1 A2
A2
+
= i2 A21 exp (ikz) ,
z
u 2 t
ikL
2 sin
, (5.29)
A2 (L) = i2 A10 kL2 L exp
2
2
Part A
Part A 5.3
o ()
e (2 )
a birefringent crystal
Pump amplitude
(arb. units)
Second-harmonic amplitude
(arb. units)
1.0
1.0
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
0.0
0.0
4
z/zsh
(5.31)
2 (, z) = 10 () tanh [10 () z] .
(5.32)
Nonlinear Optics
A2 (z, t) = i2
1
A210 t z/u 2 + u 1
u
2
1
z
A3 (z, t) = i3
exp (ik) d .
1
A10 t z/u 3 + u 1
3 u1
1
A20 t z/u 3 + u 1
3 u2
(5.33)
221 (2)
(1 ; 3 , 2 ) ,
(5.37)
k1 c2
222 (2)
2 =
(2 ; 3 , 1 ) ,
(5.38)
k2 c2
223 (2)
3 =
(5.39)
k3 c2 SFG
are the nonlinear coefficients, u 1 , u 2 , and u 3 and k1 , k2 ,
and k3 are the group velocities and the wave vectors of
the fields with frequencies 1 , 2 , and 3 , respectively,
and k = k1 + k2 k3 is the wave-vector mismatch for
the SFG process.
As long as the intensity of the sum-frequency field
remains much less than the intensities of the laser fields,
the amplitudes of the laser fields can be can be assumed to be given functions of t, A1 (z, t) = A10 (t) and
exp (ik) d.
(5.40)
(5.41)
(5.42)
where,
222 (2)
(2 ; 1 , 3 ) ,
k2 c2
223 (2)
3 =
(3 ; 1 , 2 )
k3 c2
2 =
(5.43)
(5.44)
(5.46)
Part A 5.3
Part A
where
= 42 3 |A1 |2 (k)2 .
(5.47)
Away from the phase-matching condition, the amplification of a weak signal is achieved only when the intensity
of the pump field exceeds a threshold,
I1 > Ith =
n 1 n 2 n 3 c3 (k)2
,
(2) 2
32 3 DFG 2 3
(5.48)
where we took
(2)
Part A 5.4
(5.50)
As can be seen from (5.50), optical parametric amplification preserves the phase of the signal pulse. This
property of optical parametric amplification lies at the
heart of the principle of optical parametric chirped-pulse
amplification [5.19], allowing ultrashort laser pulses to
be amplified to relativistic intensities. It also suggests
a method of efficient frequency conversion of few-cycle
field waveforms without changing the phase offset between their carrier frequency and temporal envelope,
making few-cycle laser pulses a powerful tool for the
investigation of ultrafast electron dynamics in atomic
and molecular systems.
In the nonstationary regime of optical parametric amplification, when the pump, signal, and idler
fields propagate with different group velocities, useful
and important qualitative insights into the phase relations between the pump, signal, and idler pulses can
be gained from energy and momentum conservation,
1 = 2 + 3 and k1 = k2 + k3 . These equalities dictate
the following relations between the frequency deviations
j in the pump, signal, and idler fields ( j = 1, 2, 3):
1 = 2 + 3
(5.52)
1 /u 1 = 2 /u 2 + 3 /u 3 .
(5.53)
and
(5.54)
3 = q3 1 ,
(5.55)
and
1
1
1
where q2 = (u 1
1 u 3 )/(u 2 u 3 ), q3 = 1 q2 .
In the case of a linearly chirped pump,
1 (t) = 1 t 2 /2, the phases of the signal and idler
pulses are given by m (t) = m t 2 /2, where m = qm 1 ,
m = 2, 3. With qm
1, the chirp of the signal and idler
pulses can thus considerably exceed the chirp of the
pump field.
Nonlinear Optics
(5.56)
n2 L
.
(5.58)
c
t
The resulting spectral broadening of the pulse can be
estimated in the following way:
(t) =
I0
(5.59)
= n 2 L ,
c
3 (3)
(; , , ) .
2n 20 c
(5.61)
(5.63)
NL (, z)
I0 ()
= SPM
z.
t
(5.64)
0.2
2
0.0
5
3
2
5
3
4
(arb. units)
Fig. 5.5 Self-phase-modulation-induced spectral broadening of a a sech-shaped laser pulse with an initial pulse
width of 30 fs in a fused-silica optical fiber with n 2 =
3.2 1016 cm2 /W. Curve 1 presents the input spectrum
of the pulse. The input pulse energy (2) 0.1 nJ, (3) 0.2 nJ,
and (4) 0.3 nJ
Part A 5.4
(t) = n 2 I (t) L .
(5.57)
c
Due to the time dependence of the radiation intensity within the light pulse, the nonlinear phase shift is
also time-dependent, giving rise to a generally timedependent frequency deviation:
10
Part A
pulse
(, z) 2SPM
I0 (0)
z .
02
(5.66)
Part A 5.4
q 1 2 q
+
+ |q|2 q = 0 .
2 2
(5.68)
1
( 0 )
u
1
+ 2 ( 0 )2 + . . . ,
2
(5.69)
q (, ) = sech () exp i
.
(5.70)
2
The radiation peak power required to support such a soliton is given by
P0 = |2 | /(02 ) .
(5.71)
Nonlinear Optics
Ap , 0 d ,
(5.72)
5.4.4 Self-Focusing
Self-focusing is a spatial counterpart of self-phase modulation. While SPM originates from the time-dependent
11
(5.74)
P0
z 2
1
,
f 2 (z) = 1 +
(5.78)
L df
Pcr
where P0 is the total power of the laser beam and
c2
(5.79)
16 2 n 2
is the critical power of self-focusing. The focal length
of the nonlinear lens is given by
L df
L sf =
(5.80)
1/2 .
P0
1
Pcr
Pcr =
Part A 5.4
0
(3)
where (s ; s , p , p ) is the third-order nonlinear-
12
Part A
Part A 5.4
With P0 > Pcr , the nonlinear lens leads to a beam collapse. In reality, beam collapse can be arrested by the
saturation of optical nonlinearity occurring at high field
intensities.
Beyond the paraxial approximation, the scenario of
self-focusing is much more complicated. The beam does
not collapse as a whole, as the focal length of the nonlinear lens for peripheral beams differs from the one
for paraxial beams. In the quasistationary regime, i. e.,
when the pulse duration 0 is much larger than the characteristic response time of optical nonlinearity nl , the
length of self-focusing is a function of time, giving rise
to moving foci [5.26]. In the nonstationary regime, i. e.,
on time scales less than nl , the leading edge of the
pulse experiences no focusing, but induces a nonlinear
lens that focuses the trailing edge of the pulse. As a result, the beam becomes distorted, evolving to a hornlike
pattern [5.27].
The equation of self-focusing (5.75) allows a waveguide solution [5.28], which corresponds to the regime
where the nonlinear lens exactly compensates for the
diffraction of the laser beam. This solution is, however,
unstable with respect to infinitely small fluctuations,
which either give rise to a diffraction divergence or
lead to a beam collapse. Such nonlinear waveguides
can be stabilized, as shown by Fibich and Gaeta [5.29],
by reflections of light from guiding boundaries in optical waveguides. To illustrate this regime of nonlinear
beam dynamics, we consider a cylindrical gas-filled
hollow waveguide and define nondimensional cylindrical coordinates r and z as r = R/R0 (R is the
dimensional radial coordinate and R0 is the inner radius of the hollow fiber) and z = Z/L df (Z is the
dimensional longitudinal coordinate). The radial profiles
Q (r) of light intensity distribution in the waveguide
solutions exp(iz)Q (r) to the NLSE governing
self-focusing (i. e., waveguides induced through the Kerr
nonlinearity along the z-axis on a bounded domain with
a circular symmetry) are described [28, 29] by solutions to the ordinary differential equation Q
Q + Q 3 = 0 [ = 2 /r 2 + (1/r)/r], subject to
the boundary conditions dQ (0)/ dr = 0, Q (1) = 0.
Although this model neglects the fields outside the fiber
core (e.g., radiation modes), it provides useful physical insights into the spatial self-action in hollow PCFs.
This differential equation has an infinite number of
(n)
solutions Q , n = 0, 1, 2, . . .. The Hamiltonians for
(n)
all the guided modes Q are positive, preventing the
(n)
blowup of the profiles Q in the presence of small
fluctuations, thus stabilizing the nonlinear waveguides
(0)
in a hollow fiber. The ground-state solution Q is
1, k 1
(3)
FWM
2, k 2
4, k4
3, k 3
b)
(3)
THG
, k
3, k TH
c)
3
1, k 1
(3)
CARS
2
1
CARS , kCARS
3, k 3
2, k 2
d)
(3)
DFWM
, k 1
, k 2
, k 3
, k 4
Nonlinear Optics
13
(5.83)
With no depletion of the pump fields, the SVEA equations give the following expression for the envelope of
the i-th Cartesian component of the FWM field [5.9]:
22FWM (3)
(FWM ; 1 , 2 , 3 )
kFWM c2
ikz
A1 j A2k A3l exp
2
sin kz
2
exp (FWM z) kz z ,
[AFWM (z)]i =
E j = A j exp i k j r j t + c.c. ,
(5.81)
(5.84)
where
k = k1 + k2 + k3 kFWM
+ i (1 + 2 + 3 FWM )
(5.85)
Part A 5.4
14
Part A
Part A 5.4
where j = 1, 2, 3, 4.
The phase-conjugate geometry of DFWM is shown
in Fig. 6d. In this scheme, two strong counterpropagating pump fields E1 and E2 with the same frequency
and wave vectors k1 and k2 = k1 illuminate a medium
(3)
with a cubic nonlinearity DFWM = (3) (; , , ).
The DFWM interaction of these two pump fields with
a weak signal of the same frequency and an arbitrary
wave vector k3 gives rise to a field with a frequency
that propagates in the opposite direction to the signal
beam and that is phase-conjugate of the signal field. The
phase-conjugate field generated through DFWM can be
instructively thought of as a result of scattering of the
forward pump field off the grating induced by the backward pump and the signal field or as a result of scattering
of the backward pump from the grating induced by the
forward pump and the signal field.
The nonlinear polarization responsible for phase
conjugation in DFWM is
.
(3)
(3)
(5.87)
PDFWM = 6DFWM ..E1 E2 E3 .
If the depletion of the pump fields is negligible, the
nonlinear propagation equation (5.24) for the considered DFWM process is reduced to the following two
equations for the amplitude of the signal field and its
phase-conjugate [5.1]
dA3
= i3 A3 + i4 A4 ,
(5.88)
dz
dA4
= i3 A4 i4 A3 ,
(5.89)
dz
where
12 (3)
DFWM |A1 |2 + |A2 |2 ,
(5.90)
3 =
cn
12 (3)
4 =
A1 A2 .
(5.91)
cn DFWM
By introducing
A3 = B3 exp (i3 z) ,
A4 = B4 exp (i3 z) ,
(5.92)
(5.93)
|4 | cos (|4 | L)
(5.96)
(5.97)
(5.99)
(5.100)
B1 = tA1 + rB2 ,
(5.101)
Nonlinear Optics
tA1
,
1 r 2 exp (2ikL L)
B1 =
tA1
,
1 R exp (i)
I2 / I1
(5.102)
= 0 + NL ,
(5.104)
where
3
1
2
1
(5.105)
(5.106)
(5.107)
0
0.0
0.2
0
0.6
I2 / I0
0.4
Fig. 5.7 The function F(I2 ) and the ratio I2 /I1 , appearing
on the right- and left-hand sides of (5.107), plotted as functions of the intensity I2 normalized to I0 = /(n 2 L) for
different input field intensities I1 : (1) I1 = I , (2) I1 = I ,
and (3) I1 = I , I > I > I . The circles in the Airyfunction curve show the range of input intensities where
the operation of the FabryPerot cavity is unstable
where
n = n0 + n2l
2
T + 4R
T sin
,
(5.108)
nl
Signal
BS1
M2
BS2
Output 1
n = n0
M1
nl = 0
Output 2
Part A 5.4
L
c
is the linear phase shift and
NL = 2n 2 I L
c
0 = = 2n 0
15
(5.103)
F (I2 ) =
16
Part A
Part A 5.4
(5.110)
(5.111)
(5.112)
(5.113)
1
1 exp p L .
p
(5.114)
(5.115)
where
P0 = s Beff
Beff = (2)1/2 [|g2 (s )|I0 L eff ]1/2 ,
and
g2 (s ) = ( 2 gR /2 )|=s .
With an assumption of a Lorentzian Raman gain
band, the critical pump power corresponding to the
threshold of SRS effect is given by an approximate
formula
Pcr
16Seff
,
gR L eff
(5.116)
Nonlinear Optics
17
Part A 5.4
d)
c)
Fig. 5.9ad SEM images of photonic-crystal fibers: (a) periodic- and (b) double-cladding fused-silica PCFs, (c) highindex-step birefringent PCF, and (d) hollow-core PCF
18
Part A
Part A 5.4
(5.117)
(5.118)
is then given by
k = kh 3kp k0 + + 3SPM P ,
(5.119)
1
+
A = i1 A |A|2 ,
(5.120)
t p z
+
B = i (A)3 exp (ik0 z)
t h z
(5.121)
+ 2i2 B |A|2 ,
where vp and vh are the group velocities of the pump and
third-harmonic pulses, respectively, and 1 , 2 and are
the nonlinear coefficients responsible for SPM, XPM,
and THG, respectively; and k = kh 3kp is the phase
mismatch (or the difference of propagation constants in
the guided-wave regime) in the absence of the nonlinear
phase shifts of the pump and third-harmonic fields.
Solution of (5.120) and (5.121) yields [5.24, 55]
A tp , z = A0 tp exp i SPM tp , z ,
(5.122)
z
B (th , z) = i
dz A30 th + z
0
exp i0 z + 3iSPM th + z , z
(5.123)
+ iXPM th , z , z ) ,
Nonlinear Optics
XPM th , z , z = 22
19
Third-harmonic intensity
(arb. units)
(5.124)
0.4
A0 th + z 2 dz
Third-harmonic intensity
(arb. units)
0.2
1.0
0.0
1
z
1
0
(5.125)
2
A d d ,
(5.126)
0.5
1
2
0.0
Part A 5.4
20
Part A
dth dz
B (, z) = i A
2
2
3 th + z
exp
ik0 z + 3i1 A
2 2
2
th + z
z
exp
2
z
2
dz
exp 2i2 A
10 2
2.0 x 10 4
2
1
20
10
0
th/
10 3
10 4
z
2
th + z
exp
+ ith
2
10 5
(5.127)
B(, z) i A
dz
Part A 5.4
0
2 z
exp i k0 + 31 | A|
2 (z z )
, (5.128)
+2i2 | A|
x
where (x) = exp(x 2 ) dx.
0
With z/
1 and [(z z ) ] constant, the phase
matching is controlled by the factor
2
sin2 k + + 31 A 2z
.
F (, z) =
2 2
k + + 31 A
(5.129)
10
Nonlinear Optics
Pump intensity
(arb. units)
1
0.8
0.1
0.4
0.01
0.0
350
400
450
(nm)
10 3
1300
(nm)
3
2
0
1
1
0.30
0.35
0.40
0.45
(m)
21
Part A 5.4
22
Part A
Part A 5.5
The phenomenological approach based on nonlinearoptical susceptibilities does not include nonstationary
phenomena related to a dynamic modification of the nonlinear medium by the laser field. Interaction of a resonant
laser field with a two-level system is a physically interesting and methodologically important regime where
the equations governing the evolution of the laser field
in a nonlinear medium can be solved self-consistently
with the equations of motion for the quantum system
interacting with the field. Such a self-consistent analysis reveals the existence of a remarkable regime of
nonlinear-optical interactions. A resonance laser pulse
whose amplitude and pulse width are carefully matched
to the two-level system can propagate in a two-level
medium with no absorption-induced attenuation of the
pulse amplitude, a phenomenon known as self-induced
transparency.
The interaction of laser radiation with a two-level
system is a classic problem of laser physics [5.60]. It has
been extensively studied with the use of various approximate analytical approaches and numerical procedures
over four decades, revealing important aspects of the
interaction of laser radiation with a two-level system
and, generally, lasermatter interactions. The standard
approach based on the slowly varying envelope approximation (SVEA) and rotating-wave approximation gives
MaxwellBloch equations in the case of a two-level
atom [5.60, 61]. These equations provide an adequate
description of laser radiation propagating in resonant
media within a broad range of parameters and give the
key to understanding such fundamental resonant optical phenomena as the formation of 2 solitons and
self-induced transparency [5.6063].
As shown in the classic works by McCall and
Hahn [5.62, 63], light pulses with pulse areas multiple of can propagate in a two-level medium with no
changes in their shape, while pulses with other pulse
areas tend to change their areas during the propagation in a two-level medium evolving to pulses with
areas multiple of p. Much analytical and numerical
work has been done over the past decade to extend
this SVEA result to ultrashort pulses. Eberly [5.64] has
rederived the area theorem for the case of short light
pulses, modifying this theorem to include pulse chirping and homogeneous damping. Ziolkowski et al. [5.65]
applied the finite-difference time-domain (FDTD) technique [5.66] to solve the semiclassical MaxwellBloch
equations numerically. This approach revealed several
important features of short-pulse propagation in a twolevel medium and allowed a more detailed analysis
of self-induced transparency effects. Hughes [5.67, 68]
has employed the FDTD approach to demonstrate the
possibility of generating sub-femtosecond transients in
a two-level medium.
Tarasishin et al. [5.69] have applied FDTD technique to integrate jointly the Maxwell and Schrdinger
equations for an ultrashort light pulse propagating in
a two-level medium. Below in this section, we explain
the FDTD-based algorithm [5.69] solving the Maxwell
and Schrdinger equations to model the interaction
of ultrashort laser pulses with an ensemble of twolevel atoms. FDTD simulations presented below reveal
interesting regimes of short-pulse propagation and amplification in two-level media, including the evolution of
the pulse to a 2 soliton, amplification of a single-cycle
pulse in a medium with a spatially modulated distribution of dipole moments of resonant transitions, and
amplification of chirped light pulses.
(5.130)
(5.131)
To perform this integration, we have to define the relation between the components of the electromagnetic
induction and the electromagnetic field. This can be done
through the equation for the polarization of the medium.
In our case of a two-level medium, this involves the solution of the Schrdinger equation for the wave functions
of the energy levels.
Nonlinear Optics
5.5 Ultrashort Light Pulses: Self-Induced Transparency and the Pulse Area Theorem
We will consider an ensemble of noninteracting twolevel atoms or molecules whose wave functions can be
represented as superpositions of two basis states 1 and 2:
(t) = a(x, t)1 + b(x, t)2 ,
(5.132)
where 1 and 2 are the eigenfunctions of an unperturbed system corresponding to the states with energies
E 1 and E 2 (we assume for definiteness that E 1 > E 2 ),
respectively, and a(x, t) and b(x, t) are complex coefficients. Then, the Schrdinger equation for the wave
function yields the following set of differential equations:
da(x, t)
= E 1 a(x, t) E z (x, t)b(x, t),
dt
db(x, t)
= E 2 b(x, t) E z (x, t)a(x, t),
i
dt
(5.133)
(5.134)
(5.135)
(5.136)
(5.137)
(5.138)
(5.139)
(5.140)
dr2 (x, t)
= 0r1 (x, t) + 2 (/ ) E z (x, t)r3 (x, t) ,
dt
(5.141)
dr3 (x, t)
= 2 (/ ) E z (x, t)r2 (x, t) ,
dt
(5.142)
where 0 = (E 1 E 2 )/ . Differentiating (5.141), using (5.140) for dr1 (x, t)/ dt, and taking into consideration that r1 (x, t) = Pz (x, t)/(N) = [Dz (x, t)
E z (x, t)]/(4N), we find that
d2 Dz (x, t) d2 E z (x, t)
. (5.144)
dt
dt
The FDTD approach involves difference approximation [5.66] of time and spatial derivatives involved in
(5.130) and (5.131):
t n+1/2
n+1/2
(H
Hy,i1/2 ) , (5.145)
x y,i+1/2
t n+1
n+3/2
n+1/2
n+1
(E
Hy,i+1/2 = Hy,i+1/2 +
E z,i
) , (5.146)
x z,i+1
n
Dn+1
z,i = Dz,i +
2
n+1
n1
n+1
n
E z,i
r3,i
= r3,i
+ 0.5
+ E z,i
4 0 N
n+1
n+1
n1
.
Dz,i Dn1
z,i E z,i + E z,i
(5.147)
(5.148)
Part A 5.5
23
24
Part A
(x)
2N
=
(x) sin (x),
x
ncE 0
To test the FDTD-based procedure of simulations described in Sect. 5.5.2, we model the propagation of light
pulses
(5.149)
Part A 5.5
2N
A(x, t) n A(x, t)
+
=
sin[(x, t)] .
x
c t
nc
(5.152)
t x/V
A(x, t) =
sech
,
(5.153)
42 N 2 n
+
V=
.
(5.154)
nc
c
(5.156)
N
where = 8ncE
.
0
According to (5.156), pulses whose areas (x0 ) are
multiples of propagate in a two-level medium with
no changes in their envelopes (the soliton propagation
regime). However, solitons with pulse areas equal to ,
3, 5, . . . are unstable. Thus, a pulse with an arbitrary
initial area changes its waveform propagating in a twolevel medium until its pulse area becomes multiple of
2. The characteristic spatial scale of this process is
estimated as 1 .
Numerical simulations [5.69] have been performed
for hyperbolic-secant pulses:
2
(5.155)
(x)
= sin (x),
x
2
E(x0 , t) = E 0
2 cos [(t t0 )]
.
exp[(t t0 )/T ] + exp[(t t0 )/T ]
(5.157)
The parameters of the medium and the pulse were chosen as follows: = 2E 0 /2 = 0.0565, 4Nr3 (0)
= 0.12E 0 , and = 0 (exact resonance).
In accordance with the area theorem, a pulse with
a duration T = 5.631 2/ should propagate in the twolevel medium with these specified parameters values
without any changes in its envelope. FDTD simulations [5.69] reveal no changes in the waveform and the
amplitude of such a pulse within a distance of 100
with an accuracy better than 0.1%.
Now, we examine how an arbitrary light pulse
evolves to a 2 pulse in a two-level medium. FDTD simulations were performed for the propagation of a pulse
(5.157) with a duration T = 7.0392/ in a two-level
medium with these specified parameters values. While
the amplitude of such a pulse increases by 15% as it
propagates through the medium, the duration of the
pulse decreases by a factor of about two. The pulse
areas FDTD-simulated for these values of the propagation coordinate are equal to 2.5, 2.33, 2.24, and
2.1, while the pulse areas for the same values of x calculated from (5.156) are equal to 2.5, 2.35, 2.22,
Nonlinear Optics
5.5 Ultrashort Light Pulses: Self-Induced Transparency and the Pulse Area Theorem
and 2.09, respectively. Thus, the results of FDTD simulations agree very well with the predictions of the
area theorem, indicating the adequacy of the numerical
approach.
5.5.4 Amplification
of Ultrashort Light Pulses
in a Two-Level Medium
(5.159)
where = 4N
(0)nc and (0) is the initial energy of the
pulse.
By looking at (5.150) and (5.159) and taking
into consideration that (0) = 2 E 02 = 2E 0 /(0) for
a hyperbolic-secant pulse (5.157), we find that the pulse
area can be kept equal to by modulating the spatial
distribution of the dipole moment in accordance with
0
(x) =
.
(5.160)
1 + x
Such a modulation of the spatial distribution of dipole
moments can be achieved, for example, by preliminarily
orienting molecules in the medium.
In accordance with the area theorem, the evolution
of the amplitude of a pulse propagating in an inverted
two-level medium with a spatial profile of the dipole
moment described by (5.160) is governed by
E(x) = E 0 1 + x .
(5.161)
This growth of the pulse amplitude is exactly compensated by the decrease in the dipole moment of transitions
in the two-level medium. The net effect is that the pulse
area remains constant and equal to .
Figures 5.13ad present the results of FDTD simulations for the amplification of a light pulse with a duration
T = 2/ corresponding to a single optical cycle in an
inverted two-level medium with a uniform distribution of
the dipole moment (Figs. 5.13a,b) and in an inverted twolevel medium where the spatial distribution of dipole
moments of resonant transitions is modulated in accordance with (5.160) (Figs. 5.13c,d). Simulations were
performed for the case when = 20 E 0 /2 = 0.159,
4Nr3 (0) = 0.12E 0 , and = 0 . The time in these
plots is measured from the moment when the center of
the pulse passes through the entrance boundary of the
medium.
Figures 5.13a,b display the evolution of a pulse in
an inverted uniform two-level medium and the population difference in this medium at the distances 0, 3 1
and 6 1 . These plots show that a two-level medium
with a uniform distribution of the dipole moment of
transitions cannot ensure an efficient amplification of
a transform-limited resonant pulse. Figures 5.13c,d
present the results of FDTD simulation for a pulse
Part A 5.5
Since the numerical algorithm based on the FDTD technique is intended to simulate the evolution of very short
pulses, it enables one to explore many important aspects
of short-pulse amplification in a two-level medium, providing a deeper understanding of the problems arising in
the amplification of short pulses and the ways that can
be employed to resolve these problems.
Propagation of pulses seems to provide optimal
conditions for amplification in a two-level system, since
such pulses transfer atoms (or molecules) in an initially inverted medium to the lower state. However, the
Rabi frequency increases as a light pulse propagates
through the medium and its amplitude increases due to
the amplification. Because of this change in the Rabi
frequency, the pulse now cannot transfer inverted atoms
or molecules to the lower state, but leaves some excitation in a medium, which reduces the gain and leads to
pulse lengthening due to the amplification of the pulse
trailing edge in a medium with residual population inversion. Thus, some precautions have to be taken to keep
the area of the pulse constant in the process of pulse
amplification in the medium. Below, we use the FDTD
technique to explore two possible ways to solve this
problem: (i) modulation of the spatial distribution of the
dipole moments of resonant transitions in a two-level
system and (ii) amplification of a frequency-detuned
chirped pulse.
25
26
Part A
X ( 1)
X ( 1)
3
2
1
0
1
2
3
8
6
1.0
0.5
0.0
0.5
4
t (2/)
c) E (arb. units)
1.0
8
X ( )
X ( 1)
Part A 5.5
6
3
2
1
0
1
2
3
8
t (2/)
d) r3
1.0
0.5
0.0
0.5
4
t (2/)
e) E (arb. units)
1.0
8
t (2/)
f) E (arb. units)
4
8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8
t (2/)
4
1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
t (2/)
Nonlinear Optics
5.5 Ultrashort Light Pulses: Self-Induced Transparency and the Pulse Area Theorem
1
20 E 0 2
=
2 +
,
(5.162)
2
where = 0 is the detuning of the central frequency of the pulse from the resonance.
Figures 5.13e,f present the results of FDTD simulations for a single-cycle pulse with a quadratic initial
chirp
2 cos t + kt 2 /T 2
,
(5.163)
E(0, t) = E 0
exp (t/T ) + exp (t/T )
Fig. 5.13 (a)(d) The evolution of (a), (c) the pulse waveform and (b), (d) the population difference for a single-cycle
Part A 5.5
27
28
Part A
a) E (arb. units)
b) E (arb. units)
X ( L)
X ( L)
1.0
1.0
0.5
0.5
0.0
0.0
0.5
0.5
1.0
5
c) Population
t (2/)
1.0
5
r1
r2
0.8
1
1.0
0.8
Part A 5.5
0.6
0.6
0.4
0.4
0.2
0.2
0.0
5
t (2/)
X ( L)
1
1.0
d) Population
X ( L)
r1
r2
t (2/)
0.0
5
t (2/)
Fig. 5.14 (a) Evolution of a half-cycle 2 pulse in a two-level medium: T = 0.5T0 , 2E 0 T/ = , and N/E 0 = 0.0016,
(b) Evolution of a quarter-cycle 2 pulse in a two-level medium: T = 0.25T0 , 2E 0 T/ = 2, and N/E 0 = 0.0032.
(c), (d) Evolution of the excited- and ground-state populations in a two-level medium under the action of (c) a half-cycle 2
pulse and (d) a quarter-cycle 2 pulse: (dashed line) ground-state population r1 and (solid line) excited-state population r2 .
Simulations were performed with the use of the FDTD technique [5.70]
Nonlinear Optics
29
b)
Part A 5.6
c)
science meets the physics of low-energy unamplified ultrashort pulses in the most amazing way. The advent of
photonic-crystal fibers [5.36, 37], capable of generating
supercontinuum emission with unamplified, nano- and
even sub-nanojoule femtosecond pulses, has resulted in
revolutionary changes in frequency metrology [5.7477]
opened new horizons in ultrafast science [5.78, 79] and
allowed the creation of novel wavelength-tunable and
broadband fiber-optic sources for spectroscopic [5.80]
and biomedical [5.81] applications. The rainbow of colors produced by a laser beam Fig. 5.15 has become an
optical instrument and a practical tool.
As a physical phenomenon, supercontinuum generation involves the whole catalog of classical
nonlinear-optical effects, such as self- and cross-phase
modulation, four-wave mixing, stimulated Raman scattering, solitonic phenomena and many others, which add
up to produce emission with an extremely broad spectra,
sometimes spanning over a couple of octaves. Below,
we discuss the basic physical processes contributing
to supercontinuum generation in greater detail, with
special emphasis made on self-phase modulation, fourwave mixing, and modulation instabilities (Sect. 5.6.1),
cross-phase modulation (Sect. 5.6.2), as well as the
solitonic phenomena and stimulated Raman scattering
(Sect. 5.6.3).
30
Part A
Part A 5.6
1
p + 0 p +
up
1 2
+ 2 p + 2 P ,
(5.164)
2
1
p 0 p
up
1 2
+ 2 p + 2 P ,
(5.165)
2
where P is the peak power of the pump field,
0 ( p ) is the Kerr-effect-free propagation constant
of the pump field mode (i. e., the propagation constant of the pump field corresponding to the regime
with P = 0), u p = (/|= p )1 is the group velocity of the pump pulse, 2 ( p ) = 2 /2 | = p ,
and
= (n2 p )/(cS
eff ) is the nonlinear
coefficient,
Seff =[ |F(x, y)|2 dx dy]2 / |F(x, y)|4
dx dy is the effective area of a guided mode with the
transverse field profile F(x, y).
With the propagation constant of the pump field
written as
(5.166)
p = 0 p + P ,
the mismatch of the propagation constants of the fields
involved in the FWM process is given by
FWM = p + + p + 2 p
2 p 2 + 2 P .
(5.167)
The phase matching can thus be achieved for this type
of FWM at
1/2
2 P
=
(5.168)
2 p
0.5
0.0
600
700
800
900
1000
(nm)
Fig. 5.16 Sideband generation through modulation instability in the spectrum of an ultrashort pulse transmitted
through a photonic-crystal fiber (shown in the inset). The
input energy of laser pulses is 0.5 nJ. A scanning electron
microscopy (SEM) image of the photonic-crystal fiber is
shown in the inset
Nonlinear Optics
5.6.2 Cross-Phase-Modulation-Induced
Instabilities
g (10 10cm/s)
200
2.04
0
2.02
200
400
2.00
600
0.6
0.9
1.2
1.5
1.8
(m)
Fig. 5.17 The group velocity (the solid curve and the right-
31
Experiments [5.83] were performed with highindex-step fused-silica PCFs having a cross-sectional
geometry shown in the inset to Fig. 5.17 with a core
diameter of 4.3 m. Figure 5.17 displays the group
velocity and GVD calculated for this PCF using
the polynomial expansion technique [5.84]. The standard theory of XPM-induced MI, as presented by
Agrawal [5.85], was used to analyze the main features
of this phenomenon for fundamental-wavelength and
second-harmonic femtosecond pulses of a Cr:forsterite
laser co-propagating in a PCF with the above-specified
structure. This theory predicts that stationary solutions
to slowly varying envelope approximation equations for
the pump and probe fields including dispersion up to the
second order become unstable with respect to a small
harmonic perturbation with the wave vector K and the
frequency if K has a nonzero imaginary part. The domains of this instability can be found by analyzing the
dispersion relation
(K g/2)2 h 1 (K + g/2)2 h 2 = C 2 ,
(5.169)
where
h j = 22 j 2 2 + 4 j P j g/2 j /4 ,
C = 2 2 21 22 1 2 P1 P2 ,
(5.170)
(5.171)
Part A 5.6
32
Part A
Part A 5.6
1.0
0.1
0.5
0.01
700 800
500 600 700 800
(nm)
(nm)
c) Intensity (arb. units)
d) Intensity (arb. units)
0.0
500
600
1.0
1.0
0.5
0.5
0.0
500
600
700 800
(nm)
0.0
1.0
0.5
0.5
500
600
700 800
(nm)
600
700 800
(nm)
1.0
0.0
500
0.0
500
600
700 800
(nm)
Nonlinear Optics
Optical solitons propagating in media with noninstantaneous nonlinear response experience reshaping and
continuous frequency down-shifting due to the Raman
effect phenomenon, called soliton self-frequency shift
(SSFS) [5.86, 87]. Photonic-crystal fibers substantially
enhance this nonlinear-optical process due to strong field
confinement in a small-size fiber core and the possibility
to tailor dispersion of guided modes by varying the fiber
structure. Liu et al. [5.54] have shown that 200 fs input
pulses of 1.3 m laser radiation can generate sub-100 fs
soliton pulses with a central wavelength tunable down to
1.65 m through the SSFS in a tapered PCF. Photoniccrystal fibers with the wavelength of zero group-velocity
dispersion (GVD) shifted to shorter wavelengths have
been used for the soliton frequency downshifting of
8001050 nm laser pulses [5.88, 89]. Abedin and Kubota [5.90] have employed a PCF to demonstrate a 120 nm
SSFS for 10 GHz-repetition-rate picosecond pulses. In
recent experiments [5.91, 92], PCFs with a special dispersion profile have been shown to provide an efficient
spectral transformation of chirped sub-6 fs Ti:sapphire
laser pulses through SSFS, leading to the generation
of a well-resolved solitonic spectral component cen-
33
6
A
i
(i) k (k) k A
=i
+
i
1
+
z
k!
k
0
k=2
2
R() |A (z, )| d ,
A(z, )
(5.172)
(5.173)
|F (x, y)| dx dy
4
12 + 22
1 22
sin
t
1
(5.174)
Part A 5.6
34
Part A
Part A 5.6
5.19, 5.20). PCFs of the first type consist of a fused silica core with a diameter of 1.6 m, surrounded with
two cycles of air holes (inset in Fig. 5.20a). To find
the parameters (k) for these fibers, we numerically
solved the Maxwell equations for the transverse components of the electric field in the cross section of
a PCF using a modification of the method of polynomial expansion in localized functions [5.84]. Polynomial
approximation of the frequency dependence of the propagation constant for the fundamental mode of the
PCF computed with the use of this numerical procedure with an accuracy better than 0.1% within the range
of wavelengths 5801220 nm yields the following (k)
coefficients for the central wavelength of 800 nm: (2)
0.0293 ps2 /m, (3) 9.316 105 ps3 /m, (4)
9.666 108 ps4 /m,
(5) 1.63 1010 ps5 /m,
(6) 3.07 1013 ps6 /m. For the fundamental
mode of such PCFs, the GVD, defined as D =
2c2 (2) , vanishes at z 690 nm. Fibers of the
second type are commercial NL-PM-750 PCFs (from
Crystal Fibre). The core diameter for these PCFs was
equal to 1.8 m. The parameters (k) for these PCFs
were defined as polynomial expansion coefficients for
the dispersion profile of the fundamental mode of these
fibers provided by the manufacturer. The group-velocity
dispersion for PCFs of this type vanishes at z 750 nm.
In the case studied here, the laser field at the input
of a PCF has the form of a few-cycle pulse (the upper
a) Intensity envelope (arb. units)
1.0
Input
0.002
0.0
0.000
z = 2 cm
0.4
z = 4 cm
0.4
z = 2 cm
0.004
0.000
0.0
z = 4 cm
0.004
0.0
0.000
z = 8 cm
0.4
z = 8 cm
0.004
0.0
0.000
0.004
z = 12 cm
0.4
0.0
0.4
0.0
z = 12 cm
0.000
z = 24 cm
0
1000
2000
3000
4000
5000
(fs)
0.004
0.000
z = 24 cm
0.5
0.6
0.7
0.8
0.9
1.0
1.1
1.2
(m)
Fig. 5.19 Temporal (a) and spectral (b) evolution of a laser pulse with an initial energy of 0.25 nJ and an input temporal
envelope and chirp shown in Fig. 5.1b propagating through the second-type PCF (shown in the inset)
Nonlinear Optics
35
Input
1.0
Input
1.0
0.0
0.0
z = 2 cm
0.5
0.0
z = 6 cm
0.5
0.0
z = 4 cm
0.5
0.0
z = 5 cm
0.5
0.0
2000
4000
6000
1000
8000
(fs)
0.006
0.003
0.000
0.006
0.003
0.000
0.006
0.003
0.000
0.006
0.003
0.000
0.006
0.003
0.000
0.006
0.003
0.000
0.006
0.003
0.000
0.006
0.003
0.000
0.006
0.003
0.000
0.4
Input
z = 2 cm
z = 6 cm
0.006
0.003
0.000
z = 10 cm
0.008
0.004
0.000
z = 20 cm
0.008
0.004
0.000
0.4
z = 30 cm
0.5
0.6
0.7
0.8
0.9
1.0
1.1 1.2
Wavelength (m)
2000
3000
4000
(fs)
Input
z = 1 cm
z = 2 cm
z = 3 cm
z = 4 cm
z = 5 cm
0.5
0.6
0.7
0.8
0.9
1.0
1.1 1.2
Wavelength (m)
Fig. 5.20 Temporal (a), (b) and spectral (c), (d) evolution of laser pulses with an initial energy of (a), (c) 0.15 nJ and (b),
(d) 0.5 nJ and an initial pulse width of 6 fs in the first-type PCF (shown in the inset). The input pulses are assumed to be
transform limited
Part A 5.6
z = 30 cm
0.5
0.0
z = 3 cm
0.5
0.0
z = 20 cm
0.5
0.0
z = 2 cm
0.5
0.0
z = 10 cm
0.5
0.0
z = 1 cm
0.5
0.0
36
Part A
Part A 5.6
7700
Time (1)
(fs)
7750
7800
7900
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2
1
0.1
0.0
700
750
800
850
0.1
900
950
Time (2)
(fs)
0.0
Nonlinear Optics
10
20
30
400
600
800
1000
1200
1400
(nm)
Fig. 5.22 The spectrum of supercontinuum emission produced by 820 nm pump pulses with an initial duration of
35 fs and an input power of 320 mW in a microstructure
fiber with a length of 30 cm and the cross-section structure
shown in the inset to Fig. 5.20a.
37
physics, chemistry, and biology is far from being completely realized. Similar to frequency-tunable sources
of coherent radiation, which revolutionized nonlinear
optics in its early days, allowing many delicate spectroscopic experiments, including nonlinear spectroscopic
studies, to be performed, the impressive progress of
femtosecond lasers in the 1990s has resulted in the
breakthrough of the nonlinear Raman spectroscopy to
new unexplored areas, giving rise to several elegant
new ideas and approaches, permitting more-complicated
systems and problems to be attacked, and leading to
the measurements of fundamental importance. This new
phase of nonlinear Raman spectroscopy also promoted
Part A 5.7
38
Part A
Part A 5.7
the development of new spectroscopic concepts, including time-resolved schemes, broadband spectroscopy,
polarization measurements, and CARS generalizations
based on higher-order nonlinear processes. This conceptual and technical progress achieved in the last decade
shows us some very important features of what nonlinear
Raman spectroscopy is going to be in the nearest future,
encouraging the application of new ideas, techniques,
and methods in this area of spectroscopy.
This section provides a brief introduction to the main
principles of nonlinear Raman spectroscopy, giving
a general idea of how the measurements are performed
and the spectroscopic data are extracted from the results
of these measurements. Following this plan, we will first
give a brief introduction to the basic concepts of nonlinear Raman spectroscopy. Then, we will consider various
modifications of coherent Raman four-wave mixing
(FWM) spectroscopy, including the standard CARS
scheme, stimulated Raman scattering, Raman-induced
Kerr effect, degenerate four-wave mixing (DFWM), and
coherent hyper-Raman scattering. We will also briefly
describe polarization techniques for nonlinear Raman
spectrometry and coherent ellipsometry, allowing selective investigation of multicomponent molecular and
atomic systems and permitting the sensitivity of nonlinear Raman spectrometry to be radically improved.
Finally, in the context of the growing interest in the applications of short-pulse spectroscopy for the investigation
of ultrafast processes, we will provide an introduction
to time-resolved nonlinear Raman spectroscopy.
In practical terms, to undertake a simple threecolor CARS experiment, one generally needs three
laser sources generating radiation with the frequencies meeting the requirements specified above. In the
most popular scheme of two-color CARS, where the
anti-Stokes signal is generated through the frequencymixing scheme a = 21 2 , the number of lasers
required is reduced to two. The light beams have to
be brought into coincidence in space to excite Ramanactive transitions in a medium and to generate the
anti-Stokes signal. Introducing some delay time between pumping and probing pulses, one can also perform
time-resolved CARS measurements to keep track of the
temporal dynamics of excitation in the system under
investigation.
While spontaneous Raman scattering often suffers
from the low quantum yield, which eventually results in
the loss of sensitivity, coherent Raman scattering allows
much more intense signals to be generated, thus allowing
very high sensitivities to be achieved. Coherent Raman
scattering also offers several other very important advantages that stem from the coherent character of the
signal, which is thus well collimated and generated in
a precisely known direction.
Generally, the analysis of the amplitude, phase, and
polarization of the signal resulting from a nonlinear
wave-mixing process in a Raman-resonant medium involves the calculation of the nonlinear response of the
medium in terms of the relevant nonlinear susceptibilities and the solution of Maxwell equations for the field
of the signal. Below, we restrict our brief introduction
to the theory of nonlinear Raman processes to the description of the basic notions and terminology, which
will be employed later to explain the main concepts of
various nonlinear Raman methods, including frequencyand time-domain CARS, coherent ellipsometry, and the
Raman-induced Kerr effect (RIKE).
Excitation of Raman Modes
Within the framework of a simple, but physically instructive, semiclassical model of a nonlinear medium
consisting of noninteracting molecules with a Ramanactive vibration with frequency , the interaction of
light with molecules (atoms or ions) can be described
in terms of the electronic polarizability of molecules
depending on the generalized normal coordinate Q
(e.g., defined as the distance between the nuclei in
a molecule) [5.99]:
(Q) = 0 +
Q +... ,
(5.175)
Q 0
Nonlinear Optics
(5.176)
p = (Q)E = 0 E +
QE +. . .
(5.177)
Q 0
is the dipole moment of a molecule and E is the electric
field strength. The energy of a molecule in a light field
is written as
H = pE = (Q) E 2 .
(5.178)
n 2 2 E 4 2 Pnl
= 2
,
(5.180)
c2 t 2
c t 2
where E is the field in the light wave, n is the refractive
index, c is the speed of light, and Pnl is the nonlinear
polarization of the medium.
Consider for example the CARS process, as an example of a nonlinear Raman process, assuming that it
involves plane and monochromatic waves,
E
(5.181)
39
2
N
E E 1 E 2 ,
P (3) (a ) =
4MD (, 1 2 ) Q 0
(5.183)
where
D (, 1 2 ) = 2 (1 2 )2
2i (1 2 ) ,
(5.184)
(5.185)
2
N
,
(5.186)
(3)R =
24MD (, 1 2 ) Q 0
Part A 5.7
40
Part A
and solving (5.182), we arrive at the following expression for the intensity of the CARS signal Ia :
2
Ia (3) (a ; , 1 , 2 )
2
kl
2 sin 2
II1 I2l
,
kl
(5.187)
Part A 5.7
(3)
(3)
(3)
(3)
(3)
(3)
(5.188)
(3)
(3)
(3)
(3)
(3)
(3)
(3)
(3)
(3)
(3)
(3)
(3)
(3)
(3)
(3)
(3)
(5.192)
2 2
3 a (3) 2
Pa =
c2 n a
2
32 b
P12 P2 3
2r |J|2 dr .
w30
(5.193)
1 + 2
k ik
,
k
(5.195)
k = 2k1 + k2 , k = 2k1 k2 , = 2(z f )/b is the normalized coordinate along the z-axis ( f is the coordinate
of the beam waist along the z-axis).
Equations (5.1935.195) show that the information
on the nonlinear cubic susceptibility in the CARS signal may be distorted by phase-matching effects. This
problem has been analyzed both theoretically and experimentally for different modifications of nonlinear Raman
spectroscopy [5.102]. The influence of absorption at the
wavelength of pump and probe waves and at the frequency of the FWM signal can be taken into account
by including the imaginary parts of the relevant wave
vectors. The integral in (5.194) can be calculated in an
analytical form for several particular cases, giving a clear
physical understanding of the role of phase-matching
and absorption effects in coherent FWM spectroscopy
and imaging [5.103].
5.7.2 Methods
of Nonlinear Raman Spectroscopy
In this section, we will briefly consider standard and
widely used schemes for nonlinear Raman spectroscopy
(Fig. 5.2), including coherent Raman scattering, stimulated Raman scattering, and the Raman-induced Kerr
effect and provide a brief introduction into the vast area
of DFWM.
Stimulated Raman Scattering
The idea of using stimulated Raman scattering (SRS) as
a spectroscopic technique is based on the measurement
Nonlinear Optics
k FWM
k2
FWM
k1
Object
1 k1
b)
k FWM
k0
k1
k2
Fig. 5.23a,b Broad-beam folded coherent-anti-Stokes Raman scattering: (a) beam arrangements and (b) wave-vector
diagram
41
Part A 5.7
42
Part A
Part A 5.7
by [5.99]
(3)
Tl (1 2 ) sin2 2 1122 (1 ; 2 , 1 , 2 )
2
(3)
+ 1221 (1 ; 2 , 1 , 2 ) I22
(5.196)
in the case of a linearly polarized pump (I2 is the intensity of the pump beam and is the angle between the
polarization vectors of the pump and probe waves) and
(3)R
Tc (1 2 ) 1122 (1 ; 2 , 1 , 2 )
2
(3)R
1221 (1 ; 2 , 1 , 2 ) I22
(5.197)
Nonlinear Optics
a)
Pr
Pr
b)
Pnr
e1
(5.198)
(5.199)
Part A 5.7
Polarization Properties
of the Coherent FWM Signal
When analyzing polarization properties of the FWM
signal, one has to take into account the interference of
resonant components of FWM related to various (molecular or atomic) transitions in the medium and the
x
Pnr
e2
Methods of polarization-sensitive four-photon spectroscopy provide an efficient tool for the solution of
many problems arising in the investigation of Raman
resonances. In particular, the polarization technique is
a standard method to suppress the coherent background
in CARS measurements [5.99, 106], which makes it
possible to considerably improve the sensitivity of
spectroscopic measurements [5.99] and improves the
contrast in CARS microscopy [5.115]. Polarization techniques in FWM spectroscopy can separately measure
the real and imaginary parts of the relevant third-order
nonlinear-optical susceptibility [5.133, 134], resolve
closely spaced lines in FWM spectra of molecules [5.99,
135] and atoms [5.102], and improve the contrast of
cubic-susceptibility dispersion curves near Raman resonances [5.136, 137]. Polarization methods in nonlinear
Raman spectroscopy [5.138140] help to analyze the
interference of vibrational Raman resonances with oneand two-photon electronic resonances in CARS spectra [5.141], and can be used to determine invariants of
atomic and molecular Raman and hyper-Raman scattering tensors [5.102] and to perform conformational
analysis for complex organic molecules [5.141]. A comprehensive review of polarization techniques employed
for molecular spectroscopy was provided by Akhmanov
and Koroteev [5.99].
43
44
Part A
(5.200)
(5.201)
The frequency dependencies of the FWM polarization ellipse parameters, as can be seen from (5.198) and
(5.199), provide information concerning the phase of
the resonant FWM component, allowing a broad class
of phase measurements to be performed by means of
nonlinear Raman spectroscopy.
Part A 5.7
(5.202)
(3)nr
(3)nr
(3)nr
(5.203)
Pr = 31111 1 e1 e1 e2 + e
2 (e1 e1 ) ,
(5.205)
(3)r
(3)r
= 1221 /1111
where
and e1 and e2 are the unit polarization vectors of the light fields with frequencies 1
and 2 , respectively (Fig. 5.24b).
Suppose that the pump fields with frequencies 1
and 2 are linearly polarized and their polarization vectors e1 and e2 are oriented at an angle with respect to
each other, as shown in Fig. 5.24b. The vectors Pnr and
Pr , in accordance with (5.204) and (5.205), generally
have different orientations in space making a nonzero
angle with each other. Therefore, by orienting the polarization analyzer in such a way as to suppress the
nonresonant component of the CARS signal (by setting
the angle measured from the direction perpendicular
to the vector Pnr equal to zero, see Fig. 5.24b), one can
analyze background-free CARS spectra.
In many situations, nonlinear Raman study of
Raman-active media would be simply impossible without this technique. This is the case, for example, when
the CARS signal from a resonant gas under investigation is too weak to be reliably detected against the
nonresonant CARS signal from the windows of a gas
cell. Another example is the CARS spectroscopy on
low concentrations of complex biological molecules,
when the coherent background due to solvent molecules
may be so strong that it leaves no way to detect
the CARS signal from the molecules being studied
without polarization suppression of the nonresonant
nonlinear Raman signal. Even small deviations of
the orientation of the polarization analyzer from the
background-suppression position may be crucial, leading to dramatic changes in the signal-to-noise ratio.
Another important conclusion that can be made from
(5.204) and (5.205) is that, measuring the ratio of the
CARS signals for different polarization arrangements is
a convenient way of determining the properties of the
nonlinear susceptibility tensor and, thus, characterizing
the symmetry properties of molecular transitions under
investigation.
Coherent Ellipsometry
Coherent ellipsometry, i. e., the measurement of the parameters of the polarization ellipse corresponding to the
FWM signal, is one of widely used modifications of
polarization-sensitive four-photon spectroscopy. Below,
Nonlinear Optics
(5.206)
Py = Pr ei sin(),
(5.207)
= cos() Re(P)
= sin() Im(P)
(5.208)
(5.209)
45
shape of the line observed in a nonlinear Raman spectrum was made in our consideration, which means that
this approach can be applied to a broad class of spectral lines. This procedure can be also extended to the
time domain, allowing not only the spectra but also the
time dependencies of the real and imaginary parts of the
nonlinear polarization of a medium to be reconstructed.
Finally, close molecular and atomic lines unresolvable in
amplitude nonlinear Raman spectra can be also resolved
in certain cases with the use of the phase information
stored in coherent nonlinear spectra.
|Q(t)A (t )|2 dt .
(5.212)
Part A 5.7
46
Part A
through polarization measurements and coherent ellipsometry (see the discussion above), and the level of
coherent nonresonant background can be suppressed
by means of the relevant polarization technique [see
(5.204), (5.205) and Fig. 5.24b]. In time-domain CARS,
on the other hand, the nonresonant background appears only at zero delay time between the pump and
probe pulses, having no influence on the transient signal, while the impulse-response measurements may
directly provide the information not only on the amplitude, but also on the phase of a molecular or atomic
resonance. The development of femtosecond laser systems resulted in the impressive technical and conceptual
progress of time-domain FWM spectroscopy, allowing
photochemistry processes and molecular dynamics to
be monitored in real time (see [5.109, 110, 113] for
a review).
4s k0 k1 k2 2
D
c4 ks2 k
(3) 2
eff P0 P1 P2 F2 ,
(5.213)
2
exp ibk
2
R
exp
d
bH
(1 + i ) (k ik ) H
(5.214)
Nonlinear Optics
i
.
H =
k
(k ik )
(5.215)
In the limiting case of tight focusing, when the confocal parameter b is much less than the length of the
nonlinear medium l, b l, no increase in the CARS
power can be achieved by reducing the pump-beam waist
radius because of the simultaneous decrease in the interaction length. Mathematically, this well-known result is
a consequence of the tight-focusing limit existing for the
phase-matching integral (5.214). For small phase mismatches, kl , the phase-matching integral can be
written in this limiting case as
4 2
F2 =
.
2
1 + kk
(5.216)
k 4l 2
.
k b2
(5.217)
M exp [ ( + 4 ) l]
2 kl
sinh2 l
2 + sin
2
l2,
l 2 kl 2
+
2
2
(5.218)
47
1 + 2 + 3
1
CARS
.
lopt
ln
=
(5.219)
4
With 1 2 3 4 = , (5.219) yields
ln 3
.
(5.220)
2
a3 n 2 1
where u mn is the eigenvalue of the characteristic equation for the relevant hollow-fiber mode (the mode
parameter), n is the refractive index of the fiber cladding,
and the refractive index of the gas filling the fiber core
is set equal to unity.
Plugging optical losses into the CARS enhancement
factor by substituting (5.222) into (5.221) with u n = 2.4
for the limiting eigenvalue of the E H11 mode of a hollow
fiber, we derive the following expression for the factor
of CARS enhancement in a solid-cladding hollow fiber
CARS
=
lopt
Part A 5.8
48
Part A
n2 + 1
Part A 5.8
Optical losses, which grow with decreasing inner radius a, limit the CARS enhancement, with the factor
rapidly lowering with decreasing a for small values of the fiber inner radius. The situation radically
changes in the case of a microstructure fiber. The
magnitude of optical losses for such fibers, as mentioned above, may be on the order of 13 dB/m in
the case of fibers with a hollow core diameter of
about 15 m [5.154]. In the case of small inner radii,
microstructure fibers provide much higher CARS enhancement factors than solid-core hollow fibers. The
CARS enhancement factor in hollow microstructure
fibers with the magnitude of optical losses equal to
0.1 and 0.01 cm1 starts to exceed the CARS enhancement factor in a solid-cladding hollow fiber for core
radii less than 20 and 45 m, respectively. For hollow
fibers with small core radii, the factor may be several orders of magnitude higher than the enhancement
factor .
An additional source of radiation losses in hollow fibers is related to radiation energy transfer to
higher-order waveguide modes. The efficiency of this
nonlinear-optical mode cross-talk process depends on
radiation intensity and the mismatch kc of propagation constants of waveguide modes involved in energy
exchange. Starting with the standard expression for the
propagation constants of E Hmn modes in a hollow
fiber, we arrive at the following formula for the coherence length lc = (2|kc |)1 of the mode cross-talk
process:
lc =
a2
2 2
2
,
u 2 u 21
(5.224)
Nonlinear Optics
efficiency of nonlinear-optical processes, including fourwave mixing and coherent anti-Stokes Raman scattering,
and suggesting the principle for the creation of highly
sensitive gas-phase sensors based on nonlinear spectroscopic techniques. Hollow-core microstructure fibers
have been demonstrated to allow the waveguide CARS
efficiency to be substantially increased as compared to
standard, solid-cladding hollow fibers. The theorem predicting an l 2 /a4 enhancement for a waveguide CARS
process in a hollow fiber with an inner radius a and
length l has been extended to include new solutions
offered by microstructure fibers. The maximum CARS
enhancement in a hollow microstructure fiber was shown
to scale as 2 /2 a4 with radiation wavelength , radiation losses , and the inner fiber radius, allowing CARS
efficiency to be substantially improved in such a fiber.
This 2 /2 a4 CARS enhancement factor differs from
the /a2 ratio, characterizing waveguide SRS enhancement in a hollow microstructure fiber, which is related to
the difference in the physical nature of SRS and CARS
signals and suggests different strategies for optimizing
fibers designed to enhance CARS and SRS processes.
49
Part A 5.8
1.5
1.0
0.5
0.0
400
500
600
700
800
900
(nm)
50
Part A
Part A 5.8
Nonlinear Optics
1.0
2
10
8
4
0.5
1
4
2
0
3
630
635
640
645
650
655
660 665
Dye-laser wavelength (nm)
0.0
Transmission
452
10
442
1.0
2
1
0.5
5
4
3
0
650
655
0.0
660
665
670
Dye-laser wavelength (nm)
Part A 5.8
51
52
Part A
Part A 5.8
0.1
400
500
1.0
(cm 1)
4
600
700
800
(nm)
2
0
2
2320
2330
0.5
0.0
2324
2340
(nm)
2331
2338
(cm1)
Fig. 5.28 CARS spectrum of Q-branch Raman-active vibrations of N2 molecules in the atmospheric-pressure air
filling the hollow core of the PCF. The insets show: (1) an
image of the PCF cross section; (2) the transmission spectrum of the hollow PCF designed to simultaneously support
air-guided modes of the pump, probe, and CARS signal
fields (their wavelengths are shown by vertical lines) in
the hollow core of the fiber; and (5.3) the mismatch of the
propagation constants = 21 2 a calculated for the
a = 21 a 2 CARS process in fundamental air-guided
modes of the hollow PCF with 1 , 2 , and a being the propagation constants of the Nd:YAG-laser second-harmonic
(1 ) and dye-laser (2 ) pump fields and the CARS signal
(a ) in the hollow PCF
Nonlinear Optics
The resonant CARS signal related to Q-branch vibrations of N2 in these experiments can be reliably
separated from the nonresonant part of the CARS signal originating from the PCF walls. The spectra of the
CARS signal at the output of the PCF Fig. 5.28 are
identical to the N2 Q-branch CARS spectrum of the
atmospheric air [5.99] measured in the tight-focusing
regime. In view of this finding, the CARS signal can
be completely attributed to the coherent Raman scattering in the gas filling the fiber core with no noticeable
contribution from the nonlinearity of PCF walls.
Results presented here show that large-core-area
hollow PCFs bridge the gap between standard, solidcladding hollow fibers and hollow PCFs in terms of
53
Part A 5.9
In this section, we focus on the potential of photoniccrystal fibers [5.36,37] as novel efficient sources for nonlinear spectroscopy. PCFs are unique waveguide structures allowing dispersion [5.40] and spatial field [5.181]
profiles to be engineered by modifying the design
of the fiber structure. Nonlinear-optical PCF components and novel PCF-based light sources have been
intensely used through the past few years in frequency
metrology [5.74, 77], biomedical optics [5.81], ultrafast
photonics [5.78, 79], and photochemistry [5.182].
Coherent nonlinear spectroscopy and microscopy
open up a vast area for applications of PCF light
sources and frequency shifters. Efficient frequency conversion and supercontinuum generation in PCFs have
been shown to enhance the capabilities of chirped-pulse
CARS [5.183] and coherent inverse Raman spectroscopy [5.184]. Cross-correlation frequency-resolved
optically gated CARS (XFROG CARS) has recently
been demonstrated [5.80] using specially designed
PCF frequency converters for ultrashort laser pulses.
Novel light sources based on frequency shifting in
PCFs provide a useful tool for the measurement
of second-order optical nonlinearities in organic materials [5.185] and offer interesting new options in
CARS microscopy [5.186]. Efficient spectral broadening of ultrashort pulses in PCFs with carefully
engineered dispersion profiles [5.4244] makes these
fibers ideal light sources for pumpsupercontinuum
probe time- and frequency-resolved nonlinear-optical
measurements [5.187].
Below in this section, we demonstrate applications of PCF light sources for chirped-pulse CARS
54
Part A
a)
4
1
3
Part A 5.9
2
0.1
0.01
400
500
600
700
900
800
Wavelength (nm)
For the PCFs used in experiments, the central wavelength of Cr:forsterite laser pulses falls within the range
of anomalous dispersion. Such pulses can therefore
form solitons in the fiber. High-order dispersion induces wave-matching resonances between solitons and
dispersive waves [5.21, 22], giving rise to intense blueshifted emission, observed as prominent features in the
PCF output spectra (Fig. 5.29b). The central wavelength
of this blue-shifted emission correlates well with the
wavelength where dispersive waves guided in the fundamental mode of the PCF are phase-matched with
the soliton produced by Cr:forsterite laser pulses. The
phase matching between the soliton produced by the
input laser pulse and the dispersive wave, which defines the frequency of the dominant peaks in output
Nonlinear Optics
Cr: forsterite
Laser
55
Objective
LBO
Delay line
PCF
MS fiber
Lens
Filter
Objective
Raman medium
Spectrograph
CARS Lens
signal
Part A 5.9
56
Part A
a)
(nm)
1
2
1.0
340
340
0.8
0.6
330
310
10
580
576
7 4 1
0.0
310
10 5 0
10
/100 (fs)
Phase (rad)
40
1.0
0.8
20
3
0.4
0.0
2000
c) | |
584
100
20
10
2000
t (fs)
0.5
d)
588
1000
2 5
/100 (fs)
5 10 15
/100 (fs)
(3) 2
1.0
0.8
0.6
0.4
0.2
0.0
588
584
I1
320
0.2
592
330
0.4
320
I2
b) CARS (nm)
580
576
4
2
/100 (fs)
0.1
950 1000 1050 1100
(1 2) / 2c (cm1)
Part A 5.9
0.0
600
700
800
900
(nm)
Fig. 5.32 (a) A diagram of femtosecond CARS spectroscopy with chirped pulses is shown on the left. The first
pulse (frequency 1 ) is transform limited. The second pulse
(frequency 2 ) is linearly chirped. The linear chirp of the
second pulse maps the delay time between the pulses on the
frequency, allowing the frequency difference 1 2 to be
scanned through Raman resonances by tuning the delay time . (b) The intensity of the CARS signal generated
in a noncoplanar boxcars geometry from toluene solution
as a function of the wavelength and the delay time between the second-harmonic and the blue-shifted PCF output
pulses used as a biharmonic pump. (c) A model of the Raman spectrum, including a doublet of Lorentzian lines and
a frequency-independent nonresonant background used in
the theoretical fit. (d) Theoretical fit of the XFROG CARS
spectrogram
The linear chirp defines a simple linear mapping between the instantaneous frequency of the blue-shifted
PCF output and the delay time , allowing spectral measurements to be performed by varying the delay time
between the pump pulses. Experiments were performed
with 90 fs second-harmonic pulses of the Cr:forsterite
laser (at the frequency 1 ) and the linearly chirped
pulses from the PCF (the frequency 2 ) as a biharmonic pump for the CARS spectroscopy of toluene
solution. The frequency difference 1 2 was scanned
Nonlinear Optics
57
Part A 5.9
58
Part A
Part A 5.9
reached from the ground state via two optical transitions, making the molecules of an aggregate share two
excitations, form a two-exciton band. The third-order
susceptibility (3) , responsible for four-wave mixing
(FWM) processes, can involve only one- and twoexcitons. Although three-exciton states can be reached
Intensity (arb. units)
1.0
1.0
0.5
0.5
2
0.0
500
600
700
800
(nm)
0.0
Two-exciton band
Chirped probe
One-exciton band
fs pump
Ground state
Fig. 5.33 (a) Absorption spectrum of the thiacarbocyanine film (1), spectrum of the second-harmonic output
of the Cr:forsterite laser (2), and spectrum of the blueshifted PCF output (3). (b) Diagram of the pumpprobe
nonlinear-absorption spectroscopy of exciton bands in molecular aggregates using a femtosecond (fs) pump pulse and
a broadband chirped probe field
Nonlinear Optics
measured by the pumpprobe technique for thiacarbocyanine J aggregates in a polymer film with a delay time
between the probe and pump pulses equal to (a) 100 fs,
(b) 500 fs, and (c) 1100 fs
0.1
0.0
0.1
630
700
770
b)
0.04
0.02
0.00
0.02
0.04
600
650
700
750
550
600
650
700
750
(nm)
0.03
0.02
0.01
0.00
0.01
0.02
by three transitions, they do not contribute to the thirdorder polarization as they do not have a transition dipole
that would couple them to the ground state. Timeresolved nonlinear spectroscopic methods based on (3)
processes, such as spectroscopy of nonlinear absorption, degenerate four-wave mixing, and third-harmonic
generation, have proven to be convenient and infor-
Part A 5.9
550
c)
0.03
59
560
60
Part A
Part A 5.10
clearly seen in the case of a centrosymmetric material. In this situation, the electric-dipole SHG and SFG
response from the bulk of the material vanishes. At
a surface or an interface, the bulk symmetry is broken, and electric-dipole second-order nonlinear-optical
Nonlinear Optics
Reflected
second
harmonic
Incident
beam
2
Medium 1
Surface / Interface
Medium 2
Transmitted
second
harmonic
b)
1
2
SF
2
SF
c)
Forward CARS
detection
1
Lens
Filter
Objective
CARS
Sample
1
2
Objective
CARS
nonlinear optical processes are not strictly forbidden even in a centrosymmetric medium. Beyond the
electric-dipole approximation, the second-order nonlinear signal, as can be seen from (5.65.8) and (5.14),
can be generated through the electric-quadrupole and
magnetic-dipole effects. It is generally very difficult, often impossible, to completely distinguish between the
surface and bulk contributions to the nonlinear signal.
Luckily enough, the electric-quadrupole and magneticdipole components in the nonlinear-optical response
are typically ka times less significant than the dipoleallowed part [5.9], with k = 2/ and a being a typical
size of an atom or a unit cell in a crystal. The ra(2)
tio of the surface dipole-allowed susceptibility s to
(2)
the bulk susceptibility b can be then estimated as
(2)
(2)
|s |/|b | d/(ka), where d is the thickness of the
surface layer. In reflection SHG (Fig. 5.35a), the bulk
contribution is typically generated in a subsurface layer
with a thickness of d /(2). The ratio of the surface part of the total reflected SHG signal to the bulk
contributions in this case is on the order of d 2 /a2 ,
which can be easily made much larger than unity. This
ratio can be substantially enhanced on frequency resonances or with an appropriate choice of polarization
arrangement.
A combination of the high spatial and temporal resolution with a spectral selectivity makes (2) techniques
a powerful tool for time-resolved species-selective studies of surfaces and buried interfaces (see, e.g., [5.204] for
a comprehensive review of recent results), detection and
size and shape analysis of adsorbed species, nanoparticles, and clusters on surfaces [5.205, 206], as well as
imaging and microscopy of biological species [5.207].
The sensitivity and selectivity of the (2) technique
are enhanced when the frequency of one of the laser
fields (1 in the inset to Fig. 5.35b) is tuned to a resonance with a frequency of one of the vibrations typical
of species on a surface or an interface (Fig. 5.35b).
This method of surface analysis is referred to as
sum-frequency surface vibrational spectroscopy. The
capabilities of this technique have been impressively
2
1 2
Dichroic
mirror
CARS
Filter
Lens
Epl-CARS
detection
61
Part A 5.10
62
Part A
(5.226)
Part A 5.10
(2)nr
+
(2)
s =
q
aq
,
1 q + iq
(5.229)
with (2)nr being the nonresonant quadratic susceptibility andaq , q , and q being the strength, the frequency,
and the damping constant for the q-th vibrational mode,
respectively.
When the infrared field is scanned over the frequency
of the q-th vibrational mode, the SFG signal is resonantly enhance, and its spectrum gives the spectrum of
the vibrational mode. Tunable dye lasers [5.204], optical parametric oscillators and amplifiers [5.205], or PCF
frequency shifters [5.185, 208] are employed as sources
of frequency tunable radiation, allowing selective probing of vibrational (as well as electronic and excitonic)
transitions in molecules and molecular aggregates.
Through the past few years, nonlinear-optical methods of surface spectroscopy have been extensively
involved in the rapid growth of nonlinear microscopic
techniques based on (2) and (3) processes. In particular, SHG and THG processes have proven to be convenient techniques for high-resolution three-dimensional
microscopy of biological objects [5.207,209], as well as
laser-produced plasmas and micro-explosions [5.210].
In early experimental demonstrations of SHG microscopy, grain structures and defects on the surface
of thin films were visualized using SHG in transmission [5.211] and surface monolayers were imaged by
reflection-geometry SHG microscopy [5.212]. In recent
years, progress in laser technologies and the advent of
new-generation imaging and scanning systems made it
possible to extend nonlinear microscopy techniques to
three-dimensional structures, buried objects, and biological tissues [5.207,209]. In this modification of nonlinear
microscopy, the nonlinear signal is generated in the
focal region of the laser beam in the bulk of a sample, originating from optical micro-inhomogeneities,
which break the point-group symmetry of the medium
or change phase-matching conditions. In two transverse
dimensions, the high spatial resolution of nonlinear microscopy techniques is controlled by the nonlinear nature
of the process, tightly confining the area where the signal is generated to the focal region. Resolution in the
direction of probing is achieved either due to symmetry
breaking, similar to nonlinear-optical surface-imaging
techniques, or through phase-matching modification.
In CARS microscopy [5.207,209,213,214], the nonlinear signal is resonantly enhanced when a frequency
difference between two laser fields is tuned to a Ramanactive mode of molecules under study, as described in
Sect. 5.4.8 (see also Fig. 5.35c). This makes microscopy
also species-selective as Raman resonances serve as
fingerprints of a certain type of molecules or molecular aggregates. Forward CARS and backward CARS
(also called epi-CARS) geometries have been developed (Fig. 5.35c). In the forward-CARS microscopy,
the mismatch |k| of the wave vectors involved in
wave mixing (Sect. 5.7) is typically much smaller than
2/CARS , where CARS is the wavelength of the CARS
signal. For epi-CARS, |k| is larger than 2/CARS .
The intensity of the epi-CARS signal can thus be comparable with the forward-CARS signal intensity only
for a thin sample with a sample thickness d meeting
the inequality |k|d < . CARS microscopy has been
one of the most successful recent developments in the
field of nonlinear-optical microscopy. In addition to the
high spatial resolution, this technique has a number
of other important advantages over, for example, microscopy based on spontaneous Raman scattering. In
particular, in CARS microscopy, laser beams with moderate intensities can be used, which reduces the risk of
damaging biological tissues. The anti-Stokes signal in
CARS microscopy is spectrally separated from the laser
beams and from fluorescence, as the anti-Stokes wavelength is shorter than the wavelengths of the laser beams.
In transparent materials, CARS microscopy can be used
to image tiny buried objects inside the sample in three
Nonlinear Optics
63
Part A 5.11
followed by a broad plateau of nearly constant conversion efficiency, ending up by an abrupt cutoff. The
existence of such a plateau was clearly a nonperturbative signature of the laser-atom interaction. Most of
the early work concentrated on the extension of the
plateau, i. e. the generation of harmonics of higher
frequency and shorter wavelength going progressively
from 20 nm at the end of the 1980s to 7 nm by
the middle of the 1990s [5.220224]. It was soon realized that, in contrast to the ideas promoted in the
nonlinear-optics community, the shortest wavelengths
were obtained with long-wavelength lasers. Today, harmonic spectra produced with short and intense laser
pulses extend to the water window (below the carbon
K-edge at 4.4 nm) [5.225, 226].
A breakthrough in the theoretical understanding of
high-order harmonic generation processes was initiated by Krause and coworkers [5.227] who showed
that the cutoff position in the harmonic spectrum follows the universal law Ip + 3Up , where Ip is the
ionization potential, whereas Up = e2 E 2 /4m2 , is the
ponderomotive potential, i. e. the mean kinetic energy acquired by an electron oscillating in the laser
field. Here e is the electron charge, m is its mass,
and E and are the laser electric field and its frequency, respectively. An explanation of this universal
fact in the framework of a simple semiclassical theory was found shortly afterwards [5.228, 229], and
confirmed by quantum-mechanical calculations [5.230,
231].
Progress in experimental techniques and theoretical understanding stimulated numerous studies of
harmonic generation. The influence of the laser polarization was investigated in great detail [5.232239]
(see also [5.240242] for the theory). The nonlinear
conversion process was optimized with respect to the
laser parameters [5.243, 244], and to the generating
medium [5.245250]. Finally, the spatial [5.251256]
and temporal [5.257263] properties of the radiation
were characterized and optimized. The specifications
of the harmonic emission (ultrashort pulse duration,
high brightness, good coherence) make it a unique
source of XUV radiation, used in a growing number of
applications ranging from atomic [5.264, 265] and molecular [5.266268] spectroscopy to solid-state [5.269
64
Part A
271] and plasma [5.272274] physics. Finally, it has recently been demonstrated that the low-order harmonics
are intense enough to induce nonlinear optical processes
in the XUV range [5.275277].
Almost immediately after the first observation of the
harmonic plateau, it was realized that, if the harmonics
were emitted in phase, i. e. phase-locked, the temporal structure of the radiation emitted from the medium
would consist of a train of attosecond pulses separated by half the laser period [5.278, 279]. There is
a clear analogy here with mode-locked lasers, where axLens
Gasjet
5
3
Laser
pulse
..
.
Part A 5.11
H43
H31
Experimental Method
Generating high-order harmonics is experimentally simple. A typical set up is shown in Fig. 5.36. A laser pulse
H49
6
4
2
H53
0
50
60
70
80
Energy (eV)
Fig. 5.36 Schematic representation of a typical experimental setup for high-order-harmonic generation. An intense
short pulse laser is focused into a vacuum chamber containing a gas medium. Harmonics are emitted along the laser
propagation axis. A photograph of a gas target is shown
at the bottom, on the left. A typical spectrum obtained in
neon is shown on the right. It shows a plateau ending with
a sharp cutoff
Nonlinear Optics
Intensity
Ar
10 6
Dephasing limited
Ne
4.003
10
10
Ne
He
10 8
20.18
Oxygen
K-edge
18
Ar
10 9
39.95
10 10
36
Kr
83.80
10 11
54
Xe
131.3
65
Ionisation energy Ip
He
Order
Carbon
K - edge
Absorption limited
10 12
100
30
300
Harmonic order
Fig. 5.37 Illustration of the influence of the gas species on the generation efficiency (left). The result presented at the
right has been obtained with 5 fs 800 nm laser pulses (after [5.302])
laser pulse. Rare gases are the favored species, for obvious technical reasons. In addition, some work has
been done with alkali atoms [5.296], ions [5.220, 304],
molecules [5.305308] and clusters [5.309311]. Photons are separated in energy and detected by an XUV
spectrometer, including a grating, sometimes a refocusing mirror, and a detector (electron multiplier,
microchannel plates, etc.). A typical experimental spectrum is shown in Fig. 5.36. This result has been obtained
in a gas of neon using a 100 fs 800 nm Ti:sapphire
laser [5.312]. It shows odd harmonic peaks up to the
53rd order, with a rapid decrease beyond the 49th harmonic, characteristic of the cutoff region. The spectral
range of the harmonic emission as well as the conversion efficiency depends strongly on the gas medium.
As schematically illustrated in Fig. 5.37 (top), the efficiency is highest in the heavy atoms Ar, Xe, Kr, but
the highest photon energies are obtained in He and
Ne [5.222, 245, 302]. Figure 5.37 (bottom) presents
Laser
Atom
Harmonic
emissiom
Laser
Medium
Coherent
superposition
Fig. 5.38 Illustration of the two aspects of high-orderharmonic generation. (top) Harmonic emission from
a single atom (bottom). Phase matching in the nonlinear
medium
Part A 5.11
66
Part A
V(r)
V(r)
Part A 5.11
Nonlinear Optics
v0
+ sin(t0 ) + (t t0 )v0 cos(t0 ) ,
(5.231)
67
is the laser period) are of interest for harmonic generation. When the laser field changes its sign, they come
back towards the core (at x = 0) with a certain kinetic
energy. This energy, which determines the emitted harmonic order, is proportional to the square of the slope
of the trajectory as it crosses the time axis [open circles in Fig. 5.40 (top)]. Except for the trajectory starting
at approximately 0.3T , giving the highest kinetic energy (cutoff), there are two (main) trajectories leading
the same kinetic energy. This is illustrated in Fig. 5.40
(bottom), showing the kinetic energy when the electron returns to the core (solid line), as well as the time
spent in the continuum (dashed line), as a function of
release time. As shown in the figure, for each energy,
and hence for each harmonic order, there are mainly two
Dipole strength (arb. units)
1010
1012
Distance from the nucleus (arb. units)
40
1014
60
1016
T/4
Time
T/2
1018
1
100
5
4
14
2
Intensity (10 W/cm )
5
Time in the continuum (optical cycle)
4.5
4
3.17
3
1.0
Return energy
3.5
2
3
2
1
0
0.2
0.5
2.5
2
0.3
0.0
0.4
0.5
Release time (optical cycle)
Fig. 5.40 (Top) Electron trajectories in the continuum corresponding to different release times. The laser electric field
is represented in dotted line. (Bottom) Kinetic energy (solid
line) and time in the continuum as a function of the release
time
1.5
20 10
10
20
30
40
50 60 70
(10 14 cm2/W)
Part A 5.11
80
0
68
Part A
a)
Phase-locked
generating beams
b)
Part A 5.11
19
21
23
25
31
27
29
Harmonic order
Nonlinear Optics
pol
q = q q
kq z qk1 z + q arctan(2z/b) dip . (5.233)
In this equation, the first two terms characterize the
difference in dispersion at the fundamental and q-th harmonic frequencies, mainly due to the effect of the free
electrons in the medium. The third one is a geometrical
term originating from the Gouy phase shift of a Gaussian beam with confocal parameter b across the focus.
The fourth term is the dipole phase described above.
According to the previous discussion, it is more physically correct to consider separately the contributions
from the different quantum paths, before making a (coherent) sum. The phase difference to minimize depends
on the trajectory and is given by
pol
q, j = q q
kq z qk1 z + q arctan(2z/b) j I(r, z, t) .
(5.234)
69
q (rad)
120
100
80
60
40
20
0
20
40
10
r (m)
60
40
20
0
20
40
60
10
10
z (mm)
L coh (mm)
1
0.8
0.6
0.4
0.2
0
10
z (mm)
shown in Fig. 5.43) lead to a contribution to q increasing with z, while the (normal) dispersion leads to
a term decreasing with z, usually quite small. In contrast, the contribution of the dipole phase, here due
mainly to the second trajectory, with a transition to the
cutoff for |z| 7.5 mm, first increases, then decreases
with z. In the particular case shown in Fig. 5.43, phase
matching on the propagation axis is best realized for
z 7 m, requiring the medium to be located after the
focus. In general, the situation can be rather complicated. Methods have been developed to visualize where
(in the nonlinear medium) and when (during the laser
pulse) phase matching was best realized [5.317], using
in particular three-dimensional maps representing the
local coherence length, defined as
q 1
Lq =
.
(5.235)
z
An example of such a map for the situation corresponding to Fig. 5.43 (top) is shown at the bottom of the same
Part A 5.11
We now turn to the second aspect of harmonic generation, namely the response of the whole medium. To
achieve phase matching, i. e. to ensure efficient conversion, the wave vector difference (traditionally called the
phase mismatch) between the harmonic wave and the
polarization must be minimized, so that the phase difference varies as little as possible over the mediums length.
For an incident Gaussian beam, this phase difference on
the propagation axis (z) is given by
70
Part A
Part A 5.11
eq
Esin0t
|eq |
1 Time (2/0)
0.5
0
| Eq | 2
1
300 as
0.5
1 Time (2/0)
the laser half-cycle. Phase locking between consecutive harmonics is not realized. In some conditions,
however, phase matching results in efficient generation
of only one of these contributions [Fig. 5.44 (right)]
[5.325, 326], leading to a train of attosecond pulses.
Another possibility, that makes use of the different
spatial properties of the contributing trajectories, is to
select the shortest trajectory by spatially filtering the
harmonic beam with an aperture. Finally, a spectral
filter can be used to select only the harmonics in the
cutoff region, where the electron dynamics is much simpler, with only one electron trajectory in a single-atom
response.
80
Part A
References
5.1
5.2
5.3
5.4
5.5
5.6
5.7
Part A 5
5.8
5.9
5.10
5.11
5.12
5.13
5.14
5.15
5.16
5.17
5.18
5.19
5.20
5.21
5.22
5.23
5.24
5.25
5.26
5.27
5.28
5.29
5.30
5.31
5.32
5.33
http://www.springer.com/978-0-387-95579-7
http://www.springer.com/978-0-387-30420-5